Information & Ethics Committee on Feb. 19th, 2019

[Aaron Snow]

It is incumbent upon any service that handles personally identifiable information or other sensitive information to provide the basic tools for a confidence-inspiring level of security. It's a common best practice.

For instance, today we use what is called "two-factor authentication". You enter not only a password, but also, after doing so, you need to be able to either get a code on your phone or carry around a physical key. I don't have mine on me; it's in my bag. It provides a belt-and-suspenders solution. The systems that provide that are providing an order of magnitude more security and confidence in the systems, and they are much more difficult to scam.

[Jacques Gourde, Conservative, Lévis—Lotbinière, QC]

Will the new digital solutions make departments so interconnected that they will be able to search for Canadians' information without their knowledge, or will individuals still have to consent?

Say the Canada Revenue Agency conducts an audit on someone, will the agency be able to turn to the Department of Finance for information on the individual? Could the agency obtain the source of a person's income, access to their banking data and other such information unbeknownst to the individual?

[Alex Benay]

Current laws prohibit us from doing that. They tend to be structured vertically within an institution.

Our Privacy Act also says that information is collected and used for the purpose for which it was collected.

Legally speaking, it would be very difficult to do what you just described without notifying Canadians.

When an enterprise service becomes a reality, or if we ever achieve a state of—let's call it—Utopia, where all services are available to Canadians on a single portal, we'll certainly have to consider those kinds of questions.

Currently, however, things are structured more vertically, as I said, preventing that type of data sharing and analysis.

[Jacques Gourde, Conservative, Lévis—Lotbinière, QC]

Thank you.

[The Chair, Conservative, Bob Zimmer]

You have about a minute.

[Jacques Gourde, Conservative, Lévis—Lotbinière, QC]

Thank you. That's it for me.

[The Chair, Conservative, Bob Zimmer]

You're good. Okay.

Next up, for seven minutes, is Mr. Angus.

[Charlie Angus, NDP, Timmins—James Bay, ON]

I'm interested in the question of AI. I see that it gets promoted a lot in smart governance. We can use AI to help fight climate change. I'm not making that up; I saw that. We can use AI, and it will help to mitigate natural disasters.

No offence to my colleagues on the government side, but governments love things with all the bells and whistles, and that are magic and seem to offer miracle cures.

I'm interested in the disenfranchisement of citizens through the use of AI, and the way that some end up as winners and losers in the digital and social realms. The ideas that are used in the modelling of AI will have social impacts.

How do we, in our work in committee, ensure that when we're talking about AI, there's an ethical lens that is transparent and makes sure that people are not being targeted or disenfranchised because they don't fit the algorithm?

[Alex Benay]

That's one of the burning questions we have in the digital service space writ large, not just in Canada, but frankly around the world. It is a question of values. Different countries will automate different things according to their values framework. We want to make sure that Canada automates some of its services based on our values framework. It is a continuous conversation.

Currently, the directives that we're putting in place are looking at certain elements, for example, making sure that we don't have a black box making a decision on behalf of a human. We know that the decision pattern of the algorithms as they change is an important factor, because how do we authenticate the fact that the Government of Canada is responsible for this service if it's an algorithm?

We also have directives in there where, according to the level of severity, you may have to have an internal peer review of the automation that you're working on. That's something we're considering, as well, in putting in some of these directives. It's all part of the algorithmic impact assessment tool that we talked about previously, which we've developed collaboratively around the world.

Those are some of the examples.

I would also point to data. We could have the most unbiased code ever and have bad and biased data. We've seen examples in the private sector, from recruitment tools at Amazon to other things, where the data was biased and therefore the service and algorithm became biased.

It's not just a question of the technology; it's actually a question of the data holdings that we currently have as well. Those are all things that we're looking at.

It's not a perfect solution, but it's something that's going to have to iterate quite frequently.

[Charlie Angus, NDP, Timmins—James Bay, ON]

I'm very interested in the black box issue. Corporations and vendors who want to deliver services are going to see that as proprietary technology. Is the black box going to be open to review by bureaucrats, or is it going to be an open source so that the public can see the calculations that are going into deciding how the algorithms work?

[Alex Benay]

That discussion is happening literally as we speak, so I can't answer that definitively one way or another. It is something that we want to take into consideration. There are pros and cons to both sides of this thing. We want to make sure that we have the best technology available, but at the same time, we have to respect certain values and ethics that are important in the country.

We do have instances where we have private sector code that is being held in case there is an emergency. Those are the things that we're looking at as well.

To be very transparent, this is a new space. Governing algorithms is a new space. Canada is one of the countries actually taking the lead on some of these things. The key for us will be to work iteratively with our global partners who share the same values.

[Charlie Angus, NDP, Timmins—James Bay, ON]

Something happened today in the House of Commons that I think shocked all of us was an accusation of racial profiling of young black activists who came to Parliament Hill. I have been 15 years in Parliament. I'm a white guy with white hair, and I think Parliament is very open and inclusive—the police are excellent—but we had a case of people being profiled because of their colour.

My point is that when we're looking at profiling and the black box technology—and the bureaucracy might be the ones who can best see that—the latter will be challenged by people who are being victimized by the profiling that is done. That's my question: How do we ensure that there is that kind of public input and public challenge to make sure that the algorithms do reflect the values of our society in a way is moving everyone forward?

[Alex Benay]

We have a few existing mechanisms. For example, the architecture review board, which I mentioned a few times, will be looking at automation projects as these are identified in investment plans, departments and other things. From concept, we will be looking at applying our values throughout, from procurement to deployment. The architecture review board will look at that. I think the level of transparency of our services will be very important as we move forward. Some of our directors were looking at the facts we must disclose, such as that you are talking to a robot and not a person on the other side of the chat, for example—which will be good—and that we are expecting citizens to report issues back. The dialogue will increase and become very active as we start automating. It is a pace-of-change discussion as much as it is a values discussion.

[Charlie Angus, NDP, Timmins—James Bay, ON]

Thank you for that. As a musician, I know what a polyrhythm is. I don't really know what an algorithm is, but I think it's really important. I just want to know that other people have an opportunity to give their input.

[Aaron Snow]

Perhaps I'll just note that we're used to talking about computer code as something that we understand in terms of how that code functions. We don't generally treat each other that way. Increasingly, as the code gets more complex and harder for us to comprehend ourselves, it may become more important to focus on, in addition to how the code operates, the outcomes that are actually occurring. If the outcomes are going in the wrong direction, then hold the systems responsible for the output and not just for the nature of the decision-making itself.

[Charlie Angus, NDP, Timmins—James Bay, ON]

Thank you very much.

[The Chair, Conservative, Bob Zimmer]

We'll go to Mr. Saini for one question.

[Raj Saini, Liberal, Kitchener Centre, ON]

Two.

[The Chair, Conservative, Bob Zimmer]

I thought you said one.

[Raj Saini, Liberal, Kitchener Centre, ON]

They are two small ones.

[The Chair, Conservative, Bob Zimmer]

Take your time. Go ahead.

[Raj Saini, Liberal, Kitchener Centre, ON]

I just want to talk about technology, but since Mr. Angus raised the point about AI, I'll address it this way. If we look at the advancement of AI right now, there is some speculation that in 15 to 20 years, 40% of the current job market will be displaceable. You can see how fast the private sector is moving here. In terms of the government sector—because there are going to be points, when you have a digital government architecture, where you will also need touchpoints for the private sector—how will you make sure that the technology will remain so that both systems are somewhat equal?

[Alex Benay]

That's a very good question. I would answer with a few things.

First, I think artificial intelligence and automation, like any new technology.... My colleague, Mr. Snow, spoke to a hype cycle. It's actually a thing in our sector where we see where that technology sits on the hype cycle. I still think that artificial intelligence and the concept of job replacement is, frankly, something that is very debated. It has gone from “the machines will take our jobs” to our increasingly seeing coexistence of humans and machines, and how that will actually augment the work as opposed to replace the work.

I can't speak definitively on the strategy because I still think it's very early in the hype cycle. We're seeing that the countries that do this well are the ones that not only make the investments in these technologies, but also bring a multidisciplinary lens to this discussion.

It's not only an economic growth issue; it's a service to Canadians issue, writ large. It's also how the private sector serves citizens as well. I think it will be important for us to keep an eye on the developments. Often things are done so quickly.... A service provider from a company without necessarily.... It happens so quickly we don't see it. For example, translation services are increasingly becoming very efficient on social media platforms, and it's only for the last eight months that you've actually seen an efficiency curve really spike. So, all of a sudden, that's a different conversation than we were having six months ago.

I think the key will be multidisciplinary, making sure that we don't bring departments to an issue, but bring a horizontal sort of lens to the issues and just continue that narrative and the dialogue.

[Raj Saini, Liberal, Kitchener Centre, ON]

I have a final question. As I'm sure you're well aware, in 2007 Estonia faced a cyber-attack, and one of the responses to that cyber-attack was to have scalable blockchain technology, which they found to be secure and had a strong amount of integrity for the system. Are there any thoughts about employing that here?

[Alex Benay]

I would say that we are increasingly mature on the artificial intelligence path. We're a little less mature on blockchain in the Government of Canada. It doesn't mean that we don't have places that are experimenting. We are looking at ID pilot, digital identification pilots with provinces, for example, using blockchain. We're looking at grants and contributions, I believe, at the National Research Council, using blockchain. It is probably the next area that we need to look at. I would like to say we're more mature in that space, but we're actually not.

We invested heavily in the AI space to match the investment that the government's made in AI across the country, to make sure we were ready to ingest those services. I would suspect blockchain and a few other concepts will probably be next on our radar.